Consolidate vision finetune files under finetune/vision/

Force train GPU via decorator env, not with_options

Add v5 train config (restraint data, 2 epochs, no in-train eval)

Add synthetic-only v4 train config + A100/skip-upload wiring

Inline workspace guard in train/merge so remote containers import

Wire v3 train, merge, GGUF export, and runtime selection

Fix lint and format drift in modal and evals scripts

Fix Modal profile imports for remote eval containers

Pin Modal jobs to marinaleitecabrera workspace

Add DSPy-aligned v2 vision fine-tune pipeline (T12)

Apply ruff format across the repository

Omit preprocessing_num_workers from vision LoRA config (T12)

Skip LLaMA-Factory transformers pin during Modal train (T12)

Pin transformers>=5.7 for MiniCPM-V 4.6 Modal training (T12)

Fix MiniCPM-V 4.6 processor preflight for Modal LoRA (T12)

Fix Modal vision training failures (T12)

Add Modal QLoRA vision training scaffold (T12)